SWI Compute emerges from stealth with $1bn NVIDIA order for Europe

The Euronext-listed group's AI infrastructure arm launches with 3.6 GW of secured power and a UK factory targeting 7,000 Rubin GPUs

A long data center hallway is lined with server racks displaying glowing blue and green fiber optic cables, illuminated by overhead ceiling lights and white floor strips.

SWI Compute, the newly launched compute division of Euronext Amsterdam-listed SWI Group, has emerged from stealth with a plan to convert the parent company's six-year land-and-power land bank into a network of NVIDIA-certified AI factories across Europe and the United States. The company announced it has placed approximately $1 billion in NVIDIA hardware orders, secured its first customer contracts and achieved Preferred NVIDIA Cloud Partner status as of August 2026.

The launch centres on a UK campus that will open its first phase by end-2027, initially drawing 30 MW and housing close to 7,000 NVIDIA Rubin GPUs and 3,500 Vera CPUs in NVIDIA Vera Rubin NVL72 racks. That initial phase is positioned as the first step towards a 530 MW UK campus, and towards 2.3 GW of AI factory capacity group-wide by 2029. SWI Group holds 3.6 GW of contracted power across 24 sites in Ireland, the UK, Denmark, Spain, Italy and the US, with at least one campus already leased to a major hyperscaler.

The infrastructure proposition

SWI Compute's stated differentiator is vertical integration from grid connection to GPU rack. Every site is built to NVIDIA's DSX AI Factory Platform reference architecture, which co-designs compute, networking, power and cooling as a single system. The company says the DSX design can unlock up to 40% more GPU capacity within a fixed power envelope, a claim it attributes to NVIDIA's own specification rather than independent benchmarking.

The first deployment uses NVIDIA Blackwell Ultra systems, with Vera CPU racks to follow, connected by NVIDIA Quantum-X800 InfiniBand and Spectrum-X Ethernet. Delta Computers will serve as systems integrator. SWI Compute says its initial fleet will exceed 8,000 GPUs in total. Its first named customer is a US-based AI inference platform; terms were not disclosed.

"Energy availability represents the primary bottleneck facing European AI infrastructure," said Halil Mirakhmedov, chief executive of SWI Compute. "Just two months after joining the NVIDIA Cloud Partner programme, our team secured approximately $1 billion in advanced hardware orders and executed initial client contracts."

Nico Caprez, Vice President for Global AI Infrastructure Growth at NVIDIA, said SWI Compute is "building infrastructure that is productive, durable and fungible, able to serve new customers and workloads and keep generating value for years to come."

Market and regulatory context

SWI Compute enters a European AI infrastructure market defined by two structural constraints: power scarcity and hyperscaler priority. Established cloud providers and frontier AI labs have historically absorbed available capacity on long-duration contracts, leaving AI-native startups and mid-market enterprises to compete for whatever remains. A number of well-capitalised independent AI cloud operators, including CoreWeave, Lambda Labs and local European players, are pursuing similar strategies of pre-securing power and building to NVIDIA reference architectures, making differentiation on delivery timelines and customer access increasingly important.

The European angle carries regulatory weight. Customers handling personal data under GDPR, or subject to forthcoming obligations under the EU AI Act, face genuine legal constraints on cross-border data transfers for training and inference workloads. SWI Compute's positioning as a European-jurisdiction provider, with campuses on the continent and in the UK, directly addresses those requirements, particularly for life sciences, financial services and public-sector customers.

The UK's National AI Strategy and related compute commitments have raised political expectations around sovereign AI capacity. A 530 MW campus in the UK, if delivered on schedule, would represent a material addition to domestically available AI compute. Investors will be watching near-term milestones: first GPU racks energised, named customer revenue, and whether the Q4 2027 second-campus timeline holds as hardware supply chains and grid connection queues remain under pressure across Europe.